Perishable inventory optimisation
Order too little fresh stock and the shelf runs empty. Order too much and it ends up in the bin. Our model orders each item by how long it keeps.
- Free 30-minute call
- No commitment
- You own the code
In short: A perishable inventory optimisation sets how much of each fresh item to order for each delivery, based on how long it keeps. It weighs the cost of goods thrown away against the sales lost when the shelf is empty.
The problem
On Saturday at 5 pm, the shelf for chicken salad bowls is empty. Three customers have asked for one since lunch. On Monday morning, the dairy shelf shows the opposite: twelve tubs of yoghurt reach their best-before date and go in the bin. Your ERP, the system that places your orders, ordered both with the rule it uses for tinned tomatoes.
With perishable goods, both mistakes cost money, often in the same week. Order too little, and the shelf is empty before the next delivery. Order too much, and stock you’ve already paid for expires before anyone buys it. A tin that doesn’t sell today sells next month. A salad that doesn’t sell today is food waste, and you pay twice: once to buy it and once to dispose of it.
Signs your fresh ordering runs on gut feel
Waste and empty shelves show up in the same category, often in the same week. Store managers top up the proposed orders by hand because they don’t trust them. Safety stock is the reserve you keep for days with more demand than expected. Items that keep for three days get the same safety stock rule as pasta or rice. Older packs expire at the front of the shelf while customers take fresh ones from the back. Last month you threw away less, but your shelves were empty more often. Nobody can say which of the two was worth more money.
What changes
Today, you order fresh items with a fixed rule, such as stock for two days of sales. When the bin fills up, you correct the order by hand. Our model orders each item for each delivery. It knows how long the item keeps and how many packs of each age are still on the shelf. It weighs a thrown-away pack against a lost sale in francs, item by item. Where a price reduction on the last day brings in more than it costs, the plan includes it.
What you get
We build the model on your sales, deliveries and waste records, per store and item. Your demand forecast goes in as an input. So does how far the forecast was off in the past. Before anything goes live, the model orders past weeks again, as if it had been in charge. You compare its waste and its sold-out days with what happened. So you see the effect on your own figures before you commit.
- A model written around your own perishable inventory optimisation rules.
- Results measured in the numbers you already track.
- It slots into how you already work, and you can check every answer it gives.
How it works
Forecasts, stock and decisions about your warehouse network go into one model instead of three spreadsheets. By network we mean all your warehouses, distribution centres and branches, and which site supplies which. The model plans forecasts, stock and network decisions together in a single run. So a fix in one warehouse no longer creates a new problem at another site.
Why the safety stock formula fails on fresh food
The textbook safety stock assumes that a unit you don’t sell today is still for sale tomorrow. For a tin, that’s true. For a sandwich that keeps one day, every unit of buffer you didn’t need goes in the bin that evening. So the buffer has a price that long-life goods never pay.
That price grows with the service level you ask for. Here, service level means how often the item is still on the shelf at closing time: at 95 percent, it sells out early on 5 days in 100. Each extra point needs a bigger buffer, and on fresh food much of that buffer is thrown away. Our model sets the service level per item. For that it uses two figures: what a lost sale costs you and what a wasted pack costs you. You stop paying for a large buffer on items where an empty shelf in the evening costs less than the waste.
The newsvendor model: how much to order for one day
For an item that keeps only one day, operations research has a classic answer: the newsvendor model. It’s named after a newspaper seller who buys papers each morning that are worthless by evening. Take a salad bowl you buy for 4 francs and sell for 9. A missing bowl costs you the margin, 9 minus 4, so 5 francs. A bowl left over costs you the 4 francs you paid for it.
The model adds up both losses: 5 francs plus 4 francs makes 9. A missing bowl accounts for 5 of those 9, which is about 56 percent. So you order enough to cover demand on about 56 days in 100, a little more than an average day’s sales. For bread with a wide margin and cheap ingredients, the share is higher, and you order more generously. For an expensive item with a thin margin, it’s lower, and you order lean.
Several days of shelf life: stock by age
Most fresh items keep longer than a day: yoghurt for weeks, salad for a few days, fish for two. Then today’s leftover is still for sale tomorrow, only one day older. So the model tracks stock by age: how many packs expire today, tomorrow and the day after. An order can look right in total and still produce waste, because the oldest packs expire before they sell. With shelf life inventory management by age, you see which packs will expire, and when, before you place the order.
Customers take the freshest pack
How stock leaves the shelf matters as much as the order. In the warehouse, you ship the oldest stock first: first in, first out (FIFO). At the shelf, many customers reach to the back for the pack with the latest date. Then your shop sells the newest stock first. This is called last in, first out (LIFO). Older packs expire although the total looked fine. From your sales and waste records, the model sees how often customers take the older pack and how often the fresher one. It plans every order with that.
Remaining shelf life on arrival
A pack’s shelf life starts at the producer. Yoghurt with 21 days of shelf life may arrive with 12 left, and those 12 days are all you have to sell it. One supplier delivers with 18 days left, another with 10. The model plans with the remaining shelf life on arrival, per supplier and item. You see which supplier costs you the most waste, and you go into the next negotiation with figures.
Delivery days and the weekend
Delivery days shape every order. Say the truck comes on Monday, Wednesday and Friday. Then the Friday order must last until Monday. Saturday, your busiest day, lies in between, and no delivery comes. Order on Friday as you would on Wednesday, and the shelf is empty by Saturday afternoon. The model plans each order for the days until the next delivery. So your fresh produce ordering follows your delivery calendar and your busiest days.
Stochastic optimisation and MILP
The method is stochastic optimisation, which means planning with uncertain demand. Instead of one forecast number, the model works with many possible days of sales. We build them from your forecast and from how far it has missed before. Then the model looks for the order quantities where waste and lost sales together cost the least across all of those days. That’s perishable inventory optimisation, item by item and store by store.
You order in whole crates, never 0.6 of a crate. A model that keeps such whole-crate choices is called a mixed-integer linear program (MILP). Your limits go in as rules that every plan keeps: supplier minimums, shelf space, room in the cold store. You get an order proposal per item and delivery, and you see what it costs in waste and in lost sales.
Where the problem shows up most
The problem appears wherever goods expire before they sell. Grocery stores and bakeries meet it in fruit and vegetables, dairy, bread and ready meals, delivered almost every day. Makers of consumer goods and wholesalers keep stock in their own warehouses that expires soon. Hospital pharmacies and blood services keep products that expire within days or weeks. In each case, the model weighs the cost of waste against the cost of empty shelves, in francs. That’s how food waste reduction pays for itself.
Common questions.
Something else? Ask us directly.
There’s no single right figure. For each item it follows from two amounts: what you earn on a unit you sell, and what you paid for a unit you throw away. If one more sale earns more than one wasted unit costs, some waste is the right choice. Bread is an example: you earn a lot on each loaf compared with what it costs you. Order an expensive item that earns little per unit sparingly. It then causes almost no waste and sells out now and then. Ask for zero waste on every item, and many shelves are empty by evening.
Yes, if nobody accounts for it. If a salad sold out at 3 pm, your sales figures show only what was on the shelf. They don’t show how much customers wanted. Plan from those figures, and you order too little, so it sells out early again. We mark the days an item sold out, from the time of its last sale or its closing stock. The model reads those days as “at least this much”, not as quiet days.
Daily sales per item and store, ideally with the time of the last sale. Deliveries with the remaining shelf life on arrival. Your waste records, and the order and delivery calendar for each supplier. Purchase and selling prices, plus your current demand forecast. It doesn’t need to be complete on day one. We start with the items where waste and empty shelves cost you the most.
Sometimes it can, and if that’s the case, we’ll tell you. Standard software is built for an average operation, though. Yours runs on its own rules: your contracts, your exceptions, the things you know by heart. A custom model writes those rules in, and that’s where the better plan comes from.
You don’t need either to start. We build the model from the data you’ve already got, even if it lives in three spreadsheets and one person’s head. You get a system you run yourself. We train your team until they no longer need us.
You see a first result with the prototype, before you commit to anything bigger. We start with a free 30-minute call about your operation and the model that would fit it. Then we build a working prototype and run it on your numbers.
Similar pages.
Terms on this page
Every abbreviation explained in plain words, in our article “Metrics and methods”.
- Cycle service level and fill rate
- ERPEnterprise resource planning
- MILPMixed-integer linear programming
- Stochastic optimisation
Tell us what you’re planning.
In a free 30-minute call with a specialist, you find out whether optimisation can help.
- Free, no commitment
- You talk to a specialist directly
- A clear go or no-go